Container Mass Flow Detection Using Deep Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting fallen or damaged containers in a container mass flow are inflexible and require significant adaptation to different applications, container parameters, and ambient conditions, necessitating on-site setup and posing risks to personnel during manual intervention.
Innovation Solution
A method utilizing a deep neural network to evaluate image data from a camera, trained on a wide variety of container types and ambient conditions, allowing for flexible and reliable detection of fallen or damaged containers without manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional sensors or laser scanners are used to detect fallen containers, then detection function is achieved, but the system requires significant adaptation to different container types and ambient conditions
Solution Approach 1:
The patent applies universality by using a camera-based image capture system that can detect fallen containers across multiple container types without requiring sensor adaptation. The image processing unit with trained neural network models serves as a universal detection mechanism that handles various container geometries, materials, and ambient conditions through learned patterns rather than application-specific tuning.
Solution Approach 2:
The patent employs copying by creating virtual representations of containers through image capture and processing. Instead of using physical sensors that must be adapted to each container type, the system captures optical copies (images) of containers and uses neural network models to analyze these copies, eliminating the need for physical system adaptation to different container geometries.
2Ease of manufacture
If traditional detection methods are implemented, then detection capability is provided, but on-site setup and programming require expert knowledge and time
Solution Approach 1:
The patent applies preliminary action by pre-training neural network models with extensive image data before deployment. The image processing unit is equipped with previously trained models that have learned container characteristics, fallen states, and ambient condition variations in advance. This eliminates the need for on-site training and programming, allowing rapid deployment without expert knowledge at the installation location.
Solution Approach 2:
The system applies self-service through automated image processing and detection algorithms that require no manual programming or adjustment during deployment. The neural network models automatically adapt to different scenarios through their training data, and the system operates autonomously without requiring expert intervention for setup or configuration.
3Ease of operation
If manual intervention is used to remove fallen containers, then container removal is achieved, but personnel safety is compromised due to dynamic pressure in mass flow
Solution Approach 1:
The patent applies mechanics substitution by replacing manual mechanical removal operations with an automated detection and removal system. The camera-based detection system identifies fallen containers, and the system can trigger automated removal mechanisms (such as diverters or robotic grippers) to eliminate containers from the mass flow without requiring personnel to physically intervene in the high-pressure container stream.
Solution Approach 2:
The patent uses an image processing unit with neural network models as an intermediary between container detection and removal action. This intermediary system processes visual information and translates it into control signals for automated removal mechanisms, eliminating the need for direct human intervention in the dangerous zone while maintaining safe and effective container removal operations.
4Measurement precision
If application-specific sensors are used, then detection accuracy for specific container types is improved, but the system cannot be easily transferred to other applications
Solution Approach 1:
The patent achieves universality through a camera-based system that captures visual information applicable to multiple container types. The neural network models are trained on diverse image data representing various container geometries, materials, and conditions, enabling the same system to maintain high detection accuracy across different applications without requiring sensor replacement or reconfiguration.
Data Source
AI summary
Method for detecting containers which have fallen over and/or are damaged in a container mass flow, wherein the containers in the container mass flow are transported vertically on a transporter, wherein the container mass flow is captured as an image data stream using at least one camera, and wherein the image data stream is evaluated by an image processing unit, wherein the image data stream is evaluated by the image processing unit using a deep neural network in order to detect and locate the containers which have fallen over and/or are damaged.